Re: Menus borken

[Havoc Pennington <hp redhat com>]

Alexander Larsson <alla lysator liu se> writes:
> Is anyone aware of the fact that submenus have been broken for about two
> weeks?

Nope. ;-) We have bugzilla running now though, pop it in there...

> When i mouse around in a menu with submenus i quickly get this crash:
> 
> Program received signal SIGSEGV, Segmentation fault.
> 0x40145d9b in g_type_check_instance_cast (type_instance=0x1, iface_type=19) at gtype.c:2001
> 2001	  if (!type_instance->g_class)
> (gdb) bt
> #0  0x40145d9b in g_type_check_instance_cast (type_instance=0x1, iface_type=19) at gtype.c:2001
> #1  0x40390f1a in gdk_drawable_ref (drawable=0x1) at gdkdraw.c:152
> #2  0x4039229b in gdk_event_copy (event=0xbffff140) at gdkevents.c:313
> #3  0x4012d36e in boxed_proxy_collect_value (value=0xbfffee84, nth_value=0, collect_type=0xbfffee3c,
>     collect_value=0xbfffee30) at gboxed.c:130
> #4  0x4013f285 in g_signal_emit_valist (instance=0x813bb50,
> signal_id=21, detail=0, var_args=0xbffff0ec) at gsignal.c:1636

Looks to me like what changes is that now gdk_event_copy() is called
on the event for whatever signal 21 is, and that doesn't work.

So there are two problems: 
 - it should be made not to segfault, even given current boxed behavior
 - we should not be copying the event, which is a fix to boxed type
   stuff

Havoc